Hantavirus

Statistical Sensitivity for Detection of Spatial and Temporal Patterns in Rodent Population Densities

Cheryl A. Parmenter, Terry L. Yates, Robert R. Parmenter, and Jonathan L. Dunnum University of New Mexico, Albuquerque, New Mexico, USA

A long-term monitoring program begun 1 year after the epidemic of hantavirus pulmonary syndrome in the U.S. Southwest tracked rodent density changes through time and among sites and related these changes to hantavirus infection rates in various small-mammal reservoir species and human disease outbreaks. We assessed the statistical sensitivity of the program’s field design and tested for potential biases in population estimates due to unintended deaths of rodents. Analyzing data from two sites in New Mexico from 1994 to 1998, we found that for many species of Peromyscus, Reithrodontomys, Neotoma, Dipodomys, and Perognathus, the monitoring program detected species-specific spatial and temporal differences in rodent densities; trap- related deaths did not significantly affect long-term population estimates. The program also detected a short-term increase in rodent densities in the winter of 1997-98, demonstrating its usefulness in identifying conditions conducive to increased risk for human disease. We analyzed the statistical capabilities of the desert grasses (Sporobolus spp. and Bouteloua long-term rodent monitoring program begun in eriopoda). The second site (Placitas), a pinyon- 1994 to detect spatial and temporal changes in juniper woodland site located in Sandoval County rodent densities and determine if the low death in the foothills of the Sandia Mountains, Cibola rates at all study sites resulted in biased (under- National Forest, approximately 30 km north of estimated) rodent population density estimates. Albuquerque, New Mexico, USA (35º 16.7'N, 106º We also examined a short-term subset of the data 18.6'W, elevation 1,830 m), was dominated by (mid-1997 to early 1998) to test whether the pinyon pine (Pinus edulis), one- juniper, and program design could statistically detect a sudden blue grama grass (Bouteloua gracilis). rodent increase in density that may precede a hantavirus pulmonary syndrome outbreak. Statistical Methods The overall experimental design and field Study Sites sampling procedures are described by Mills et al. We selected two monitoring sites in New (this issue, pp. 95-101). The rodent density estimates Mexico for analysis because they provided the we analyzed were based on mark-recapture longest period of field sampling, the greatest trapping data from three permanently marked range in rodent species richness, and the largest trapping webs at each site (1,2). We used difference in types. The first, a desert monthly trapping data collected from August site 90 km south of Albuquerque, on 1994 to February 1998. Trapping and rodent the Sevilleta National Wildlife Refuge, Socorro handling methods were described by Parmenter County, New Mexico, USA (34º 21.3'N, 106º et al. (3), and safety procedures during animal 53.1'W, elevation 1,465 m), was dominated by one- processing followed Mills et al. (4). Blood samples seed juniper (Juniperus monosperma), honey collected from Peromyscus maniculatus were ( glandulosa), and various analyzed by the Centers for Disease Control and Prevention for Sin Nombre virus (SNV). Rodent Address for correspondence: Cheryl A. Parmenter, Depart- densities were calculated by the program ment of Biology, The University of New Mexico, 167 Castetter Hall, Albuquerque, NM 87131-1091, USA; fax: 505-277-0304; e- DISTANCE (2). Each dataset was analyzed by mail: [email protected]. three models (uniform, half-normal, and hazard)

Emerging Infectious Diseases118 Vol. 5, No. 1, January–February 1999 Hantavirus

with three possible model adjustments (cosine, ern Biology. These offspring were included for the polynomial, and hermite). Akaike’s information duration of the expected lifespan on each study criterion was then used to select the model that site. Thus, if a pregnant female N. albigula died best fit the particular dataset (2). The three web during the study, we would add two offspring density estimates for each trapping period were (the mean litter size for this species in New then partitioned into “proportional” densities Mexico) to the MNA estimates for the full three representing each species, on the basis of the trapping periods of their life expectancy. This relative proportion of each in the total web process created species-specific hypothetical MNA sample. These species-specific densities (in values that were either equal to or greater than numbers of mice per hectare) were analyzed by a the observed MNA values. The two MNA datasets repeated measures analysis of variance were then compared by RMANOVA. (RMANOVA) to test for differences between sites and through time and for site and time Spatial Differences in Rodent Densities interactions. If significant F values were observed, RMANOVA successfully distinguished ro- we conducted within-trapping-period tests to dent densities between sites for a number of detect differences among site means; we used species (Table 1). Ord’s kangaroo rat and the Fisher’s least-significant-difference method. The effect of low death rates among rodents on estimates of population size was analyzed by Table 1. Repeated-measures analysis of variance testing for differences among representative rodent minimum number alive (MNA) methods (5). For population densitiesa, Sevilleta National Wildlife Refuge this analysis, we calculated the “observed” MNA and Placitas, 1994-1998 values for each species during each sampling Species Source DF F value p period on the basis of actual field data, which Dipodomys Site 1 7.52 0.0517 ordii Error (Site) 4 included occasional trap deaths of rodents. We (Ord’s kangaroo Time 35 2.19 0.0007 then constructed a hypothetical MNA, assuming rat) Time x Site 35 2.21 0.0006 that no trap deaths occurred in the sampled Error (Time) 140 populations. The hypothetical death-free MNAs Perognathus Site 1 25.69 0.0071 were computed by extending the projected life flavescens Error (Site) 4 span of each animal that died in the trap. The (Plains pocket Time 35 2.91 0.0001 mouse) Time x Site 35 2.89 0.0001 length of the extensions differed by species and Error (Time) 140 site and was determined by the mean number of Peromyscus Site 1 0.28 0.6233 trapping periods during which each species maniculatus Error (Site) 4 would normally have been present on each site (Deer mouse) Time 35 2.80 0.0001 (this figure was based on the lifespans of all other Time x Site 35 1.18 0.2513 mice of that species that did not die in the traps Error (Time) 140 or during handling). For example, if Neotoma Peromyscus Site 1 0.11 0.7609 leucopus Error (Site) 4 albigula at the Sevilleta National Wildlife Refuge (White-footed Time 35 7.25 0.0001 site, Web 1, had a mean lifespan of three mouse) Time x Site 35 4.56 0.0001 trapping periods (i.e., its initial capture period Error (Time) 140 [time zero] plus three additional sampling periods), Peromyscus truei Site 1 1.98 0.2328 and a mouse died in a trap on its first capture, we (Pinyon mouse) Error (Site) 4 would add one mouse to the observed MNA Time 35 3.77 0.0001 Time x Site 35 3.29 0.0001 values for three additional trapping periods to Error (Time) 140 arrive at the hypothetical MNA values. If the Neotoma Site 1 6.65 0.0614 mouse died during its second trapping period, we albigula Error (Site) 4 would add two trapping periods to the MNA (White-throated Time 35 1.53 0.0431 estimates, and so on. In addition, if the dead mouse wood rat) Time x Site 35 1.65 0.0221 was pregnant or lactating, we increased the Error (Time) 140 hypothetical MNA by the average number of Reithrodontomys Site 1 0.34 0.5906 offspring that would have been produced; mean megalotis Error (Site) 4 (Western harvest Time 35 4.04 0.0001 numbers of offspring for each species were mouse) Time x Site 35 3.57 0.0001 determined from specimen databases at the Error (Time) 140 University of New Mexico’s Museum of Southwest- aDensity=number of mice per hectare.

Vol. 5, No. 1, January–February 1999119 Emerging Infectious Diseases Hantavirus

Plains pocket mouse (Heteromyidae) were much C more abundant at the Sevilleta National Wildlife Refuge site than at Placitas (Figure 1A, B) and had greater statistical differences in the RMANOVA results (Table 1). In contrast, other rodent species (Muridae) had no overall differences by site (Table 1) and usually had similar densities, except for occasional episodes (Figures 1C-G). Although we observed no overall effect of site on D density for these species, Fisher’s least- significant-difference methods showed significant differences between sites during certain periods (Figures 1C,D,F,G), demonstrating that within a particular species, intersite differences could be discerned in both long-term sequences and during episodic, site-specific population irruptions.

Temporal Changes in Rodent Densities E The analyses also detected changes in rodent densities through time in all species examined (Table 1). Several species with generally low densities (e.g., the harvest mouse [Figure 1D], the white-footed mouse [Figure 1E], and the deer mouse [Figure 1G]) occasionally became locally extinct but periodically recolonized the sites. Other species (e.g., the pinyon mouse [Figure 1F] F and the white-throated wood rat [Figure 1C]) were found consistently on both sites, although their densities fluctuated considerably. In all cases, RMANOVA found significant differences in these temporal patterns.

G A

Figure 1. Mean densities of rodents at the Sevilleta B National Wildlife Refuge (solid circles) and Placitas (open circles) study sites. Asterisks indicate significantly different means between sites (Fisher’s least-significant-difference tests, p < 0.05). A. Ord’s kangaroo rat (Dipodomys ordii). B. Plains pocket mouse (Perognathus flavescens). C. White-throated wood rat (Neotoma albigula). D. Western harvest mouse (Reithrodontomys megalotis). E. White-footed mouse (Peromyscus leucopus). F. Pinyon mouse (P. truei). G. Deer mouse (Peromyscus maniculatus).

Emerging Infectious Diseases120 Vol. 5, No. 1, January–February 1999 Hantavirus

Short-Term Rodent Population Increases Therefore, the monitoring program was capable To determine the capability of the analyses to of showing sudden, short-term increases in show statistically significant short-term increases rodent densities that may precede a disease in rodent densities (e.g., a rodent population outbreak in humans. “explosion”), we selected May 1997 to February Blood tests to determine the presence of SNV 1998, a period characterized by a rodent density in deer mice showed generally low infection rates increase in some species (Figures 1A-G). We then (Figure 2), with a maximum of only one mouse tested the datasets for this period alone; testing positive per trapping period at Placitas, RMANOVA results indicated significant in- and none at Sevilleta. The SNV-positive rodents creases in densities for the Plains pocket mouse, were detected during periods of moderate deer mouse, white-footed mouse, pinyon mouse, abundance in 1994 and 1995 but not in the early and Western harvest mouse (Table 2; Figure 1 stage of the population increase during the B,D-G) during the winter of 1997 to 1998. winter of 1997-98.

SNV Table 2. Short-term repeated-measures analysis of variance for differences among representative rodent population densities, Sevilleta National Wildlife Refuge and Placitas , May 1997 to February 1998 Species Source DF F value p Dipodomys Site 1 5.75 0.0745 ordii Error (Site) 4 (Ord’s Time 8 2.13 0.0616 kangaroo rat) Time x Site 8 2.13 0.0616 Error (Time) 32 Figure 2. Seroprevalence of Sin Nombre virus (SNV) in the deer mouse populations at Sevilleta National Perognathus Site 1 27.09 0.0065 flavescens Error (Site) 4 Wildlife Refuge and Placitas study sites. (Plains pocket Time 8 4.61 0.0008 mouse) Time x Site 8 4.61 0.0008 Error (Time) 32 Low Death Rates among Rodents and Peromyscus Site 1 2.70 0.1759 Population Estimates maniculatus Error (Site) 4 Four species had sufficient sample sizes for (Deer mouse) Time 35 2.45 0.0432 the MNA analyses, which were based on 28 Time x Site 35 1.57 0.1868 trapping periods between August 1994 and Error (Time) 140 January 1997 and included 7,024 rodent Peromyscus Site 1 2.25 0.2084 captures (Table 3). During field sampling, trap leucopus Error (Site) 4 death rates were generally lower than 10% (3). A (White-footed Time 8 7.83 0.0001 mouse) Time x Site 8 5.79 0.0001 breakdown of the number of new animals, Error (Time) 32 recaptured resident animals, and trap deaths Peromyscus Site 1 5.09 0.0870 indicated that immigration and reproduction truei Error (Site) 4 rates were consistently higher than trap death (Pinyon Time 8 7.09 0.0001 rates (Figure 3A-D). Species showing territorial mouse) Time x Site 8 4.94 0.0005 behavior (e.g., Merriam’s kangaroo rat [Figure Error (Time) 32 3A] and the white-throated wood rat [Figure 3C]) Neotoma Site 1 1.06 0.3618 had higher ratios of residents to immigrants albigula Error (Site) 4 than species without strongly defended territories. (White-throated Time 8 1.17 0.3443 wood rat) Time x Site 8 0.82 0.5944 In constructing the hypothetical MNA Error (Time) 32 values, we used the following mean life Reithrodontomys Site 1 4.58 0.0990 expectancy values (number of trapping periods megalotis Error (Site) 4 after initial capture): D. merriami = 2.16; (Western Time 8 4.67 0.0007 P. flavescens = 1.11; P. truei = 0.68; N. albigula = harvest mouse) Time x Site 8 1.64 0.1530 2.75. For female rodents of reproductive age, the Error (Time) 32 following mean numbers of offspring were used Density = numbers of mice per hectare. (on the basis of University of New Mexico’s

Vol. 5, No. 1, January–February 1999121 Emerging Infectious Diseases Hantavirus

A A

B

B

C

C

D

D

Figure 4. Mean minimum number alive (MNA) values of “observed” (solid circles) and “hypothetical” (open circles) for the following: A. Merriam’s kangaroo rat (Dipodomys merriami) B. Plains pocket mouse (Perognathus flavescens) C. White-throated wood rat (Neotoma albigula) at the Sevilleta National Wildlife Refuge study site Figure 3. Composition of sampled populations at D. Pinyon mouse (Peromyscus truei) at the Placitas the Sevilleta National Wildlife Refuge study site. study site. The number of immigrants greatly exceeds Values are means of three replicate trapping sites; incidental trap deaths. A. Merriam’s kangaroo rat error bars represent one standard deviation. (Dipodomys merriami). B. Plains pocket mouse (Perognathus flavescens). C. White-throated wood rat (Neotoma albigula). D. Pinyon mouse (Peromyscus truei).

Emerging Infectious Diseases122 Vol. 5, No. 1, January–February 1999 Hantavirus

Table 3. Repeated-measures analysis of variance for and did not return, despite the higher densities differences between actual minimum number alive observed in this species at these sites. Clearly, (MNA) estimates and hypothetical MNAs (zero deaths from trapping), New Mexico, 1994 to 1997 the population dynamics observed at this time Species Source DF F value p were not equivalent to the rodent “outbreak” of Dipodomys MNA Treatment 1 0.03 0.8672 1993. Continued monitoring of these populations merriami will be needed to determine the extent and (Merriam’s Time 27 21.19 0.0001 importance of this apparent trend in rodent kangaroo Time x MNA 27 0.30 0.9997 population ecology. rat) Error (Time) 108 The statistical sensitivity demonstrated in Perognathus MNA treatment 1 0.08 0.7858 this study is critical to the success of field flavescens Error (MNA) 4 monitoring programs, particularly those that (Plains Time 27 7.99 0.0001 function as early warning systems to alert pocket Time x MNA 27 0.01 1.0000 health-care workers and researchers of impend- mouse) Error (Time) 108 ing outbreaks of disease (7,8). In the case of SNV, Peromyscus MNA treatment 1 0.03 0.8661 the monitoring program serves as both an early truei Error (MNA) 4 warning system and as a research database to (Pinyon Time 27 6.24 0.0001 mouse) Time x MNA 27 0.13 1.0000 which numerous environmental variables and Error (Time) 108 the prevalence of SNV infections in rodents can be correlated. Neotoma MNA treatment 1 0.57 0.4914 albigula Error (MNA) 4 In addition to the need for statistical power (White- Time 27 11.82 0.0001 in distinguishing spatial and temporal patterns, throated Time x MNA 27 0.15 1.0000 the monitoring program must provide study site wood rat) Error (Time) 108 population estimates that can be directly compared. Standardization of techniques used by collaborating research groups ensures such Museum of Southwestern Biology data): comparability. In these hantavirus studies, the D. merriami = 2.03; P. flavescens = 2.75; P. truei use of trapping webs and distance sampling = 3.46; N. albigula = 1.83. When analyzed with theory (2) also allows for direct comparisons of RMANOVA, observed MNA values were not rodent densities among species and across different from hypothetical (Table 3). While widely varying ecosystems. Trapping grids, and highly significant temporal differences in MNA their associated population estimators, often values were observed (Figures 4A-D), no produce results that may be suitable for local treatment-by-time interactions were produced, studies (internally consistent within a single which demonstrated that the low death rates experiment), but cannot be compared across during the monitoring program did not affect ecosystem types and taxa because of site-specific rodent population estimates. characteristics or species-specific assumptions of capture probabilities. Trapping webs and Conclusions distance sampling density estimators, however, Our analyses indicated that the field have produced reasonably accurate density experimental statistical design was sufficiently estimates in both a computer simulation study sensitive to detect a range of differences in (9) and a field study (10). The accuracy of densities of rodent species across study sites and trapping webs and grids in estimating rodent through time. Even relatively moderate levels of densities is being evaluated more fully at the increases were detectable by our methods, Sevilleta National Wildlife Refuge site. although they were far less dramatic than those All these methods require an appropriate observed during the 1993 SNV outbreak (6). sampling design and an understanding of the While rodent densities of certain species basic biology of the rodents. For example, the significantly increased during this study (1994 to Plains pocket mouse has MNA values of four 1998), the maximum densities were considerably animals per month for the months of November, lower than those observed at the Sevilleta 1994 to March 1995 (Figure 4B). In contrast, its National Wildlife Refuge site in 1993 during the density for the same period is zero (Figure 1B), SNV outbreak (6). In addition, seroprevalence in which indicates that no animals were captured deer mice dropped to zero in 1996 (Figure 1G) during field sampling. The species’ habit of

Vol. 5, No. 1, January–February 1999123 Emerging Infectious Diseases Hantavirus

becoming inactive and remaining in under- New Mexico, the U.S. Centers for Disease Control and ground burrows during cold winter weather Prevention, the Indian Health Service, and the State of New Mexico Health Department. Additional support was provided accounts for this discrepancy. Thus, in ecologic by the Sevilleta Long-Term Ecological Research Program and perhaps disease transmission terms, (NSF Grant DEB-9411976), the Museum of Southwestern density data accurately reflect which species of Biology, and the National Institutes of Health (Grant 1 PO1 rodents are active on a site, whereas MNA data AI39780). show the combined survivorship of active and Cheryl A. Parmenter is a research associate in the Uni- inactive species. versity of New Mexico’s Museum of Southwestern Biol- A potential source of estimation bias in ogy and data manager for the New Mexico Hantavirus rodent monitoring programs is the inadvertent Monitoring Team. influence of trapping and handling of rodents during field sampling. Capturing, anesthetizing, References measuring, and collecting blood and saliva 1. Anderson DR, Burnham KP, White GC, Otis DL. samples traumatizes small animals and may Density estimation of small-mammal populations using affect future trapping success, which, in turn, a trapping web and distance sampling methods. Ecology could bias the accuracy of the density or 1983;64:674-80. population estimators. Previous studies have 2. Buckland ST, Anderson DR, Burnham KP, Laake JL. Distance sampling. Estimating abundance of biological shown various effects of trapping and handling populations. New York: Chapman and Hall, 1993. on rodent body mass, ability to trap rodents in 3. Parmenter CA, Yates TL, Parmenter RR, Mills JN, the future, and survival (3,11,12-17), but none Childs JE, Campbell ML, et al. Small mammal survival have addressed the effect of low death rates on and trapability in mark-recapture monitoring programs long-term population trends. While precautions for hantavirus. J Wildl Dis 1998;34:1-12. 4. Mills JN, Yates TL, Childs JE, Parmenter RR, Ksiazek are taken to ensure survival of sampled animals, TG, Rollin PE, et al. Guidelines for working with occasionally a few die during sampling, rodents potentially infected with hantavirus. Journal of especially those with a lower tolerance to the Mammalogy 1995;76:716-22. physical, physiologic, and psychologic stress of 5. Krebbs CJ. Demographic changes in fluctuating being captured and handled. Chronic loss of populations of Microtus californicus. Ecological Monographs 1966;36:239-73. study animals from trapping or handling could 6. Parmenter RR, Brunt JA, Moore DL, Ernest SM. The underestimate their densities when compared to hantavirus epidemic in the Southwest: rodent those of “natural” or undisturbed populations population dynamics and the implications for nearby. Results of our study indicate that death transmission of Hantavirus-associated adult respiratory rates from trapping at these sites had no distress syndrome (HARDS) in the Four Corners region. Report to the Federal Centers for Disease significant effect on long-term rodent population Control and Prevention, July 1993. Publication No. 41. estimates. Sevilleta Long-Term Ecological Research Program The existing network of rodent population (LTER) 1993:1-45. study sites seems successful in identifying local 7. Mills JN, Ellis BA, McKee KT Jr, Calderon GE, Maiztegui species-specific fluctuations in densities. Data JI, Nelson GO, et al. A longitudinal study of Junin virus activity in the rodent reservoir of Argentine hemorrhagic from these sites can be used in addressing fever. Am J Trop Med Hyg 1992;47:749-63. hypotheses on the relationships among environ- 8. Mills JN, Ellis BA, Childs JE, McKee KT Jr, Maiztegui mental factors, rodent abundance, and SNV JI, Peters CJ, et al. Prevalence of infection with the infections, as well as in providing an early Junin virus in rodent populations in the epidemic area warning for potential rodent population explo- of Argentine hemorrhagic fever. Am J Trop Med Hyg 1994;51:554-62. sions that may increase the risk for other 9. Wilson KR, Anderson DR. Evaluation of a density hantavirus pulmonary syndrome outbreaks in estimator on a trapping web and distance sampling the southwestern United States. theory. Ecology 1985;66:1185-94. 10. Parmenter RR, MacMahon JA, Anderson DR. Animal density estimation using a trapping web design: field Acknowledgments validation experiments. Ecology 1989;70:169-79. We thank the U.S. Fish and Wildlife Service (Sevilleta National Wildlife Refuge) and the U.S. Forest Service (Cibola 11. Swann DE, Kuenzi AJ, Morrison ML, DeStefano S. National Forest) for their cooperation in the study. Effects of sampling blood on survival of small mammals. This research was supported by a collaborative Journal of Mammalogy 1997;78:908-13. agreement between the Department of Biology, University of

Emerging Infectious Diseases124 Vol. 5, No. 1, January–February 1999 Hantavirus

12. Slade NA, Iskjaer C. Daily variation in body mass of 15. Kaufman GA, Kaufman DW. Changes in body mass free-living rodents and its significance for mass-based related to capture in the prairie deer mouse population dynamics. Journal of Mammalogy (Peromyscus maniculatus). Journal of Mammalogy 1990;71:357-63. 1994;75:681-91. 13. Wood MD, Slade NA. Comparison of ear-tagging and 16. Douglass RJ, Van Horn R, Coffin KW, Zanto SN. toe-clipping in prairie voles, Microtus ochrogaster. Hantavirus in Montana deer mouse populations: Journal of Mammalogy 1990;71:252-5. Preliminary results. J Wildl Dis 1996;23:527-30. 14. Slade NA. Loss of body mass associated with capture of 17. Vanblankenstein T, Botzler RG. Effect of ectoparasite Sigmodon and Microtus from northeastern . removal procedures on recapture of Microtus Journal of Mammalogy 1991;72:171-6. californicus. J Wildl Dis 1996;32:714-5.

Vol. 5, No. 1, January–February 1999125 Emerging Infectious Diseases